Representing Sequences in Description Logics

Haym Hirsh; Daniel Kudenko

Representing Sequences in Description Logics

Computer Science

Research output: Contribution to conference › Paper › peer-review

Abstract

Representing and manipulating sequences in description logics (DLs) has typically been achieved through the use of new sequence-specific operators or by relying on host-language functions. This paper shows that it is not necessary to add additional features to a DL to handle sequences, and instead describes an approach for dealing with sequences as first-class entities directly within a DL without the need for extensions or extra-linguistic constructs. The key idea is to represent sequences using suffix trees, then represent the resulting trees in a DL using traditional (tractable) concept and role operators. This approach supports the representation of a variety of information about a sequence, such as the locations and numbers of occurrences of all subsequences of the sequence. Moreover, subsequence testing and pattern matching reduce to subsumption checking in this representation, while computing the least common subsumer of two terms supports the application of inductive learning to sequences.

Original language	Undefined/Unknown
Pages	384-389
Publication status	Published - 1997

Access to Document

http://portal.acm.org/citation.cfm?id=1867465

Cite this

@conference{8f2c40ec7e86450a84582566f960d313,

title = "Representing Sequences in Description Logics",

abstract = "Representing and manipulating sequences in description logics (DLs) has typically been achieved through the use of new sequence-specific operators or by relying on host-language functions. This paper shows that it is not necessary to add additional features to a DL to handle sequences, and instead describes an approach for dealing with sequences as first-class entities directly within a DL without the need for extensions or extra-linguistic constructs. The key idea is to represent sequences using suffix trees, then represent the resulting trees in a DL using traditional (tractable) concept and role operators. This approach supports the representation of a variety of information about a sequence, such as the locations and numbers of occurrences of all subsequences of the sequence. Moreover, subsequence testing and pattern matching reduce to subsumption checking in this representation, while computing the least common subsumer of two terms supports the application of inductive learning to sequences.",

author = "Haym Hirsh and Daniel Kudenko",

year = "1997",

language = "Undefined/Unknown",

pages = "384--389",

}

TY - CONF

T1 - Representing Sequences in Description Logics

AU - Hirsh, Haym

AU - Kudenko, Daniel

PY - 1997

Y1 - 1997

N2 - Representing and manipulating sequences in description logics (DLs) has typically been achieved through the use of new sequence-specific operators or by relying on host-language functions. This paper shows that it is not necessary to add additional features to a DL to handle sequences, and instead describes an approach for dealing with sequences as first-class entities directly within a DL without the need for extensions or extra-linguistic constructs. The key idea is to represent sequences using suffix trees, then represent the resulting trees in a DL using traditional (tractable) concept and role operators. This approach supports the representation of a variety of information about a sequence, such as the locations and numbers of occurrences of all subsequences of the sequence. Moreover, subsequence testing and pattern matching reduce to subsumption checking in this representation, while computing the least common subsumer of two terms supports the application of inductive learning to sequences.

AB - Representing and manipulating sequences in description logics (DLs) has typically been achieved through the use of new sequence-specific operators or by relying on host-language functions. This paper shows that it is not necessary to add additional features to a DL to handle sequences, and instead describes an approach for dealing with sequences as first-class entities directly within a DL without the need for extensions or extra-linguistic constructs. The key idea is to represent sequences using suffix trees, then represent the resulting trees in a DL using traditional (tractable) concept and role operators. This approach supports the representation of a variety of information about a sequence, such as the locations and numbers of occurrences of all subsequences of the sequence. Moreover, subsequence testing and pattern matching reduce to subsumption checking in this representation, while computing the least common subsumer of two terms supports the application of inductive learning to sequences.

M3 - Paper

SP - 384

EP - 389

ER -